Land Use Change Detection Based On ZY-3 Images

The use of land resources is linked to sustainable development of national economy and management.Land use change detection has been regarded as a significant research in the world.With the development of remote sensing satellite,it has become the most economic mean to supervise land use change.With the comparision of high-resolution satellite images,traditioanal image processing technology of remote sensing exists certain drawbacks.At the moment,we must carve a new path to monitor land use changes through remote sensing images,promote the application high-resolution images,and design suitable means of changes monitoring.Through multiresolution segmentation,this paper's research is based on ZY-3 images in different periods in Licang Distinct,Qingdao City.According to the optimal segmentation scale of feature category,the scale layer can be built and relevant feature category can be extracted.In that way,the category results can be supervised in distinct periods and every change of feature category can be recorded.This work is meant to support the research in land change and macroeconomic adjustment.The main researches are as follows.(1)Based on multiresolution segmentation,Canny is chosen to apply in the multriresolution segmentation.The result shows that operator extraction edge Canny contributes to the segmentation precesion.For instance,the boundary segmentation of the road is more approximated to the real boundary.(2)To improved the accuracy of image classification when using fuzzy classification method for image classification,it is useful for abundant explore the weights of all band and heterogeneity factors,also with fully analyze the spectral information,texture information,and shape semantics of each object.Find the best combination of features to improve the accuracy of the classification results.The result of image segmentation shows coefficient Kappa is 0.82 and 0.85,Indicating that the classification results are highly consistent and basically consistent with the actual features.(3)According to the segmentation results,the research applys conjoint analysis.Field counter and language Python define the change type,accumulate transformation of each feature,analyze the reason of the changes,and estimate the results through variation error matrix.Through analysis,the overall detection accuracy of the change detection results was 90%,and the Kappa coefficient was 0.78.The results prove the test result is high reliable.This change estimation hardly depends on manually-tested threshold value or subjective factors.Instead,it reveals the real changes of feauture catogory,which can contribute to relevant official department.